doit_estimate_w: Estimate kernel widths for DoIt

In sieste/doit: Bayesian Computation using Design of Experiments-Based Interpolation Technique

Description

Estimate the optimal kernel bandwith for the approximation by minimising the weighted mean squared cross validation error.

Usage

doit_estimate_w(design, w_0 = NULL, optim_control = NULL)

Arguments

design	A data frame of design points and corresponding function evaluations. Must contain a column named 'f' with function values. The other columns are treated as design points.
w_0	Starting point for the optimisation routine. Either NULL (default) in wwhich case an educated guess is made, or otherwise a vector of the same dimension as the parameter space.
optim_control	A list passed to 'optimize' (for 1d problems) or 'optim' (if the parameter has dimension 2 or more)

Value

A vector of optimal kernel widths.

Examples

design = data.frame(x=rnorm(10), y=rnorm(10))
design$f = with(design, exp(-0.5*(x+y)^2))
doit_estimate_w(design)

sieste/doit documentation built on May 9, 2019, 4:10 p.m.